Still reviewing your team's VODs in a spreadsheet?
ClutchRecon turns a Valorant VOD into event by event analysis, with the proof behind every number. Player tracking and team dashboards included.
Coaches: what's the one thing your spreadsheet can't show you? 👇
#Valorant#VCT
3 years ago I started playing Valorant and got so obsessed I took a (very expensive) coaching plan in Konpeki's IRP program.
I still remember sitting there watching him scrub through my VODs by hand, pulling patterns out one clip at a time. Coming from a data science background, I couldn't unsee it, half of what he was doing was painful grunt work a machine could do in seconds. The other half, turning those patterns into "okay, here's what you actually fix," that's the part only a coach can do.
What started as a passion project, collecting my own and my friends' gameplay and training models on it, is now ClutchRecon, the first product from ClutchLabs AI.
It's a 15-stage AI vision pipeline that pulls signals straight out of raw gameplay video. No Riot API, no match data, just the video. Coaches or Staff define what metrics they need and we integrate it into their account. Everything comes out structured, with the signals right there in front of you.
So coaches and analysts can get back to what they love, instead of burning hours scrubbing footage and noting timestamps.
we just kicked off a pilot with the head coach of one of the best teams in India, and we're already talking to a few more. here's the first ugly demo of what it can do.
#VALORANTEsports
Day 4
Ran HUD detection on ~2k Valorant clips to generate training data for the next model (minimap player position tracking) on the cloud via S3.
Then shipped dev tooling: set up bidirectional sync between my journal repo and project workspace with a post-merge git hook that auto-syncs Claude Code skills and agents on every pull across workspaces and devices.
Infrastructure usually feels invisible, but it's the difference between shipping fast and drowning in coordination overhead.
#buildinpublic #machinelearning #devops
Day 3
Today I locked down the scope for ClutchLabs V1: a 12-block ensemble AI pipeline that answers "Why am I inconsistent?" for Valorant players, which other API-based stat trackers cannot.
Then built the dev system to automate the crew of agents and skills to actually ship it:
- 11 Claude Code skills (YOLO, CNN, OCR, pipelines, etc.)
- 7 agents for ML research, debugging, code review, EOD daily tweet and journal logs
-MCPs for GitHub + Notion + X wired up
#buildinpublic #machinelearning #valorant #esportsdata
Day 2 of building in public.
90% of days are just debugging, setting up CUDA dependencies for GPU training on EC2, configuring ECS, or staring at CloudWatch logs.
Still a long way to go.
Vision → https://t.co/KiwTz6Xj1g
#valorant#esportsdata#gaminganalytics
Day 1 of building in public.
Building analytics to recognize player behavior nuances in-game. Currently working on event extraction via computer vision. Here's a glimpse of progress in HUD detection so far.
Vision → https://t.co/E5bWyjOlWU
@Kuvera_In
I'm a new investor, i'm starting my MF journey with your app.However, when i'm registering my bank account details,even though my ifsc code is correct,your applicarion is giving invalid ifsc code error. My ifsc code is HDFC0005763. Can you please resolve this issue.